# 4704 A Hybrid Monte Carlo System M Ronan

- Slides: 20

4/7/04 A Hybrid Monte Carlo System M. Ronan LBNL (Berkeley) A Hybrid Monte Carlo System 1

Outline 4/7/04 Ideal and realistic detector simulations Detector response simulations Fast Monte Carlo (FMC) GISMO Hybrid Monte Carlo studies Using reconstructed tracks and clusters. A Hybrid Monte Carlo System 1

4/7/04 Perfect detector simulation A perfect detector would precisely measure all of the final state particles produced in an e+e- interaction. Since the Electro. Weak and Higgs bosons are color neutral there should be little color flow from one boson to another. Pythia fragmentation, used extensively in American LC studies, suggests that W and Z bosons would be easily separated by a perfect detector, as shown. One still needs to tune jet finding algorithms and overall jet reconstruction to optimize LC physics signals. However, real detectors are not perfect. A Hybrid Monte Carlo System e+e- -> WW and ZZ events only. 1

Ideal Detector Simulations 4/7/04 The LCD Fast Monte Carlo (FMC) providess a rather idealistic simulation of the American Large and Small detectors. American FMC-based LC studies to date have used this optimistic simulation in which W and Z bosons would be easily separated, as shown. However, real detectors probably will not be this good. A Hybrid Monte Carlo System 1

Realistic Detector Simulations 4/7/04 One should choose detector simulations carefully. Some detector models can be much more realistic than others, but are often less flexible. Simple detector models: Smear MC Particle jets. The required detector performance for reconstructing jets is often given as d. E/E(jet) = 30%/sqrt(E) The resulting separation of WW and ZZ events is shown in comparison to what was achieved by the LEP detectors. Ideal detector models: Fast Monte Carlo's (FMC's) Realistic detector models: e. g. Sim. Det and Quick. Sim Full detector simulation: LCD Framework packages. A Hybrid Monte Carlo System 1

4/7/04 Hybrid Monte Carlo System A Hybrid MC system for studying detector designs. Hybrid MC “Reconstructed” Jets - Use reconstructed charged tracks and clusters with an energy flow algoritm. - Possibly add missing tracks and clusters from FMC simulation, - or any other missing particles from MC Particle information. Java Classes: Hybrid. Model. Parameters Hybrid. Monte. Carlo (--> Driver) Fast Monte Carlo (--> Processor) Hybrid. Monte. Carlo. Processor Hybrid. Selector Useage parameters = new Optimistic. Hybrid. Model. Parameters(0. 100, 0. 95); // pt. Min, cos. Th model. A = new Hybrid. Model("Model. A", parameters); // Model. A: model. A. set. Use. Reconstruction(true, false); // Use charged but no neutral reconstruction, model. A. set. Add. FMCCharged(true); model. A. set. Add. FMCNeutral(true); // add missed FMC tracks and clusters model. A. set. Add. MCCharged(true); model. A. set. Add. MCNeutral(true); // and MC particles. Hybrid. Monte. Carlo hmc. A = new Hybrid. Monte. Carlo(model. A); add(hmc. A); model. B = new Hybrid. Model("Model. B", parameters); // Model. B: model. B. set. Use. Reconstruction(true, true); // Use charged and neutral reconstruction, . . . A Hybrid Monte Carlo System 1

-- Measurement errors 4/7/04 The Hybrid Monte Carlo can also be used to study detector response to different types of particles. As the HMC processor reads the lists of reconstructed tracks and clusters, comparisons can be made with the original MC Particle. The HMC Selector is invoked for each particle as the list of reconstructed objects is being filled. Similarly, if the option is chosen, the Selector will be invoked for all non-reconstructed FMC tracks and clusters, and any other remaining MC Particles before being added to a final “Mixed Particle” list. A Hybrid Monte Carlo System Reconstructed total energy measurement errors: (a. ) pion momentum measurement error, (b. ) photon EM calorimeter measurement error, (c. ) neutral hadron EM+HAD total energy measurement error, (d. ) same plotted separately for neutrons and Klong's. . 1

-- Handelling Gismo simulated events 4/7/04 Gismo has been in use by the American LC Detector (LCD) simulation effort since Keystone (1998), but it has real problems. For example, Gismo often adds confusing incomplete MC details that makes it hard to obtain the information needed to understand detector simulation and reconstruction effects. In this case a pion interacts somewhere and generated a shower but no final particle emerges. One has to write code to find the final particles elsewhere in the documentation. A Hybrid Monte Carlo System 1

-- Model comparisons 4/7/04 One can use the Hybrid Monte Carlo to compare different detector models by varying aspects of the HMC Models. For example, to study different Calorimeter designs, one can set Model. A to use reconstructed tracks and FMC neutrals, and Model. B to use reconstructed tracks and neutral clusters, with both models using perfect energy flow reconstruction (i. e. Cheating). In these figures one can verify that the two models add up the same number of particles, and the same total energy. A Hybrid Monte Carlo System Figure (a. ) Numbers of reconstructed tracks and clusters, FMC and MC particles added for Model. A; (b. ) same for Model. B; (c. ) Total number of particles; (d. ) Total energy “reconstructed. ” 1

Reconstruction of WW and ZZ Events 4/7/04 We can also compare Hybrid Monte Carlo detector simulations to ideal and realistic simulations. As shown above, FMC jet reconstruction is quite idealistic compared to more realistic models which suggest that a jet energy resolution 30% should be achieveable. From HMC simulations with reconstructed tracks and neutrals one finds that Model. A using FMC neutrals is somewhat more realistic, while Model. B using the current American Large (LD) detector Calorimeter design has a considerably more pessimistic performance. A Hybrid Monte Carlo System Comparison of Ideal FMC, Smeared and Hybrid Model A/B simulations of WW and ZZ reconstruction. 1

JAS 3 - As an AIDA Ntuple analysis platform 4/7/04 JAS 3 is a general purpose, open-source data analysis tool with a highly modular structure, GUI, scripting and many other features. AIDA provides abstract interfaces for histograms, ntuples, fitters, etc. One can create an analysis job to read in standard I/O files, perform reconstruction and write out AIDA ntuples. And then use JAS 3 as an analysis framework using scripts or the Java language to analyze the tuples, create other tuples, do fits and style final plots. A Hybrid Monte Carlo System 1

Detector Performance Requirements ● Momentum resolution dp. T/p. T < 1 x 10 -4 * p. T [Ge. V] ● Tracking 2 -hit separation < 2 mm ● Timing d. T < 2 nsec ● Track-cluster matching < 1 mm ● Jet energy resolution d. Ejet/Ejet < 30% / sqrt(Ejet) ● Good W/Z separation ● . . . A Hybrid Monte Carlo System 4/7/04 1

4/7/04 Hybrid Monte Carlo Studies Using a Hybrid MC system to study detector effects. W mass fits FMC – Ideal detector simulation: Perfect tracking and cluster efficiency Perfect energy flow reconstruction Smeared MC Particle Jets Includes hadronization effects. 30%/sqrt(E) - LC design goal 60%/sqrt(E) - ALEPH reference Hybrid MC “Reconstructed” Jets - Use reconstructed charged tracks and clusters with an energy flow algoritm. - Add missing tracks and clusters from FMC simulation or MC information. Hybrid A - Use reconstructed tracks Hybrid B - Use reconstructed tracks and clusters with perfect energy flow. A Hybrid Monte Carlo System Ideal FMC W mass resolution of 2. 5 Ge. V is somewhat better than the 3. 0 Ge. V resolution obtained for the LC design goal of 30%/sqrt(E). Hybrid MC W mass resolutions of 2. 9 and 3. 0 Ge. V are obtained using reconstructed charged tracks and FMC neutrals, and reconstructed charged tracks and neutrals, respectively. 1

-- LCD Tracking Simulation 4/7/04 Geant 4 Detector Simulation Provides detector hits LCD Analysis Modules: Hit smearing TPC Pattern recognition Calorimeter clustering Event display. . . LCIO JAS histograms AIDA tuples Detector: ldmar 01 Hits: TPC (cyan), Inner trackers (cyan) EM Cal (blue) Tracks (red) Clusters (green) A Hybrid Monte Carlo System 1

4/7/04 -- Tracking studies We can use the Hybrid Monte Carlo to measure the significance of improvements in detector design or in reconstruction algorithms. For example: The original LCD TPC 3 D tracking code was written in late 1998 in preparation for the Sitges LCWS. It was refered to as V 0. 8. Various improvements in the tracking were made prior to Snowmass in 2001 and the Jeju LCWS meeting, resulting in V 1. 2. Running the HMC with the two different versions of the code allows a quantitative measure of the improvement. Note: In these comparisons, the charged tracking efficiency was improved in V 1. 2. The HMC allows investigation of resulting unassociated neutral calorimeter cluster. A Hybrid Monte Carlo System Reconstructed Jet-jet W mass using reconstructed charged tracks and FMC neutrals. 1

-- Total event energy measurement 4/7/04 In comparing results of ideal and realistic detector simulations with fully reconstructed simulated events or with the output of Hybrid Monte Carlo simulations, it is quite informative to study the total event energy measurement. Here the total energy for WW and ZZ events is displayed. One sees the main peak at the CM energy and a secondary peak from events where one of the Z's has decayed into neutrinos. The long lower-energy tail is due to lost neutrinos, to energy leakage from the calorimetry and from acceptance effects. The resolution of the high energy edge is sensitive to the overall event reconstruction whereby tracks and clusters can be missed, or double counted. In these GISMO simulations, an additional small flat tail arises from confusion in parsing the MC tracking information. A Hybrid Monte Carlo System Reconstructed total event energy in WW and ZZ events. 1

-- Reconstructed W mass 4/7/04 In studying W/Z separation, one can measure the low and high mass tails in W reconstruction. Here, W jet-jet mass reconstruction is shown for two Hybrid models: One (A) using reconstructed charged tracks and FMC neutrals, and the other (B) using all reconstructed particles. At this stage in the development of reconstruction algorithms, tracks and clusters that were missed are added from FMC simulations, and perfect energy flow is used. In addition, the study uses a Cluster. Cheater to find calorimeter clusters. One sees that the W mass resolution is significantly degraded by the poor sampling of the present Large detector (LD) calorimetry. Low mass tails would effect the W and Z reconstruction efficiency. A Hybrid Monte Carlo System Reconstructed W mass using reconstructed charged tracks and “Cluster. Cheater” neutrals. 1

Summary ● ● 4/7/04 Ideal detector simulations are useful for initial physics studies. TPC response simulation and track reconstruction in hand. Calorimeter cluster finding and energy flow algorithms need to be developed. Realistic detector simulations are possible within the Hybrid Monte Carlo framework. A Hybrid Monte Carlo System 1

4/7/04 -- Ideal W/Z separation What does this mean for LC physics studies ? Reconstructed Jet-jet W mass using FMC smeared tracks and neutrals. A Hybrid Monte Carlo System 1

-- A more realistic detector simulation 4/7/04 What does this mean for LC physics studies ? Reconstructed Jet-jet W mass using reconstructed charged tracks and FMC neutrals. A Hybrid Monte Carlo System 1